Tech companies are no longer waiting for utilities to power their AI ambitions; they're building their own power plants alongside massive data centers. Meta's groundbreaking C$13 billion investment in Alberta, Canada, announced in July 2026, pairs a gigawatt-scale AI campus with a purpose-built 932 megawatt (MW) natural gas generating facility, illustrating a fundamental shift in how hyperscalers approach infrastructure. This "bring your own power" model addresses one of the most pressing challenges in AI deployment: the enormous and unpredictable energy demands of GPU clusters that concentrate far more power in each rack than traditional cloud computing.

What's Driving the Data Center Power Crisis?

The explosion of artificial intelligence workloads has created an energy bottleneck that utilities alone cannot solve quickly enough. GPU clusters used for training and running large language models consume electricity at scales that dwarf conventional data centers. A single modern GPU rack can draw hundreds of kilowatts, and a hyperscale AI facility operating at gigawatt capacity requires the equivalent power output of a small city. Rather than navigate years of utility planning and grid upgrades, tech giants are taking matters into their own hands.

Meta's Sturgeon County facility in Alberta exemplifies this strategy. The campus will initially operate at 1 gigawatt (GW) of power capacity but could eventually scale to 1.8 GW, placing it among the largest data center developments outside the United States. The company coordinated the data center with a dedicated generating station, grid upgrades, and long-term natural gas agreements as a single integrated project rather than treating power as an afterthought.

How Are Companies Solving the Power Supply Challenge?

• Dedicated Power Plants: Meta partnered with Pembina Pipeline and Morgan Stanley Infrastructure Partners to build the Greenlight Electricity Centre, a combined-cycle natural gas facility scheduled to begin service in the second half of 2030. The plant will supply 932 MW of dedicated capacity under a long-term tolling agreement, where Meta covers capacity payments and operational costs.
• Industrial Location Strategy: Rather than building near cities, Meta selected a site within Alberta's Industrial Heartland, an established development zone northeast of Edmonton that already contains pipelines, energy facilities, and petrochemical plants. This location provides access to existing infrastructure and natural gas transportation networks that would be impossible to replicate at a greenfield site.
• Water-Efficient Cooling Systems: The facility employs closed-loop liquid cooling supported by dry cooling technology that consumes no water during normal operation. This addresses environmental concerns about AI data centers' water consumption while leveraging Alberta's cool climate to improve heat-rejection efficiency.
• Phased Construction Approach: Rather than delivering the entire gigawatt at once, Meta will build the campus in phases. The Greenlight power plant has already been permitted for potential expansion to 1,864 MW, closely matching Meta's ability to expand the data center to 1.8 GW, suggesting both projects are planned with a second major construction phase in mind.

Why Are Tech Companies Partnering With Nuclear and Energy Leaders?

Beyond Meta's natural gas approach, the broader tech industry is exploring nuclear power as a long-term solution to AI's energy demands. Microsoft and Nvidia announced a strategic "AI for nuclear" partnership in March 2026, intended to streamline the permitting, design, and operations of nuclear power plants. The companies believe their artificial intelligence tools can fast-track ambitious nuclear technology builds and reduce red tape, redundancies, and engineering delays without compromising safety.

The partnership reflects a recognition that traditional nuclear development timelines are incompatible with the speed at which AI infrastructure is expanding. Microsoft's generative AI tools are already showing results in the nuclear sector. Aalo Atomics has saved $80 million annually by using Microsoft's generative AI for permitting, reducing its permitting process by 92 percent. Southern Nuclear incorporated Microsoft Copilot across its nuclear reactor fleet, while Idaho National Laboratory deployed AI to automate the crafting of engineering and safety analysis reports.

"To break this infrastructure bottleneck, we need to move away from highly customized engineering towards repeatable, reference-based delivery while maintaining regulatory standards and engineering accountability," Microsoft stated in announcing the partnership.

Microsoft, in partnership announcement with Nvidia

The AI tools envision by Microsoft and Nvidia would handle document drafting and gap analyses, gauge progress with 4D and 5D simulations that track time and costs, and detect operations anomalies early using AI-powered sensors and digital twin technology. This approach could significantly accelerate nuclear reactor deployment, which has historically suffered from construction delays and cost overruns.

What Does This Mean for the Future of AI Infrastructure?

Meta's Canadian project represents more than a single data center investment; it signals a new model where data centers are paired with dedicated power generation, influencing regional energy and infrastructure policies. The company is committing approximately C$60 million to improvements involving local roads, water systems, and other community infrastructure, demonstrating that hyperscalers are now thinking of themselves as energy developers as much as technology companies.

The construction timeline alone underscores the scale of this undertaking. Approximately 3,000 construction workers are expected on site at peak activity, with more than 300 permanent employees operating the campus after completion. The work extends far beyond data halls to include multiple computing buildings, substations, utility yards, backup power systems, administrative space, security facilities, and miles of internal electrical, mechanical, and fiber infrastructure.

This infrastructure-first approach is becoming the standard for hyperscale AI deployment. Rather than selecting a building site and applying for an ordinary utility connection, companies like Meta are spending years coordinating data centers with purpose-built generating stations and grid upgrades. The result is a more predictable, controllable energy supply that can scale with AI's explosive power demands, even if it requires companies to become energy infrastructure developers in the process.